Why are complex LLM arguments as persuasive as simple ones?
Standard persuasion research predicts that simpler, easier-to-read arguments persuade better. But LLM-generated text breaks this rule—it's measurably more complex yet equally convincing. What explains this reversal?
Standard findings in persuasion research, going back to Fluency and Disfluency studies, predict that lower cognitive effort to process an argument leads to higher persuasion. Easier-to-read text earns more agreement; complexity erodes it. The classic Carrasco-Farré finding on viral misinformation supports this — reduced cognitive effort correlates with higher virality.
The LLM persuasion study tested whether LLMs follow this rule and found they do not. LLM-generated arguments scored significantly higher on grammatical complexity (mean 13.26 vs 12.16, p<.001) and on lexical complexity measured by perplexity (111.39 vs 102.69, p<.001). They were harder to read by both standard measures. And yet they were equally persuasive to human-written arguments.
This overturns the lower-effort-equals-more-persuasion assumption for LLM-generated text. One available interpretation aligns with Kanuri et al.'s finding that higher cognitive processing on social media can promote engagement: greater complexity may signal substance and importance, prompting deeper engagement that increases persuasion. Another interpretation flows from Cognitive Surrender — when readers face complex AI-generated text, they may treat the complexity itself as a credibility signal and defer to it without genuinely processing the reasoning.
Either interpretation undermines the design assumption that simpler language is universally more persuasive. The relationship between cognitive effort and persuasion is mediated by the source attribution and engagement mode of the reader. When the source is plausibly authoritative and the reader is in a deferential posture, complexity may help rather than hurt — and LLMs systematically produce text in that combination.
Inquiring lines that read this note 16
This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.
What determines AI's persuasive power and how can it be detected or mitigated?- Why does renaming the entity change how compelling the argument feels?
- Does persuasiveness increase when LLMs argue for claims that are actually true?
- How does source attribution change the complexity-persuasion relationship?
- Does cognitive complexity strengthen or weaken persuasive impact on audiences?
- Does argument quality in textbooks differ from persuasive effectiveness in practice?
- When does analytical persuasion work better than emotional persuasion?
- How do fallacy susceptibilities relate to LLM persuasiveness in debates?
- Why does personal authenticity matter more for human persuasion than LLM?
- Why do LLMs persuade through logical appeals but humans through emotion?
Related papers in this collection 8
Papers most semantically related to this note, ranked by cosine similarity in the embedding space.
- Large Language Models are as persuasive as humans, but how? About the cognitive effort and moral-emotional language of LLM arguments
- Do LLMs Change Their Minds Like Humans? Diagnosing Human--LLM Divergence in Single-Turn Persuasion Judgments
- Exploring the Role of Prior Beliefs for Argument Persuasion
- A meta-analysis of the persuasive power of large language models
- Evaluating the Capabilities of LLMs for Persuasive Dialogue
- When Large Language Models are More Persuasive Than Incentivized Humans, and Why
- AI Argues Differently: Distinct Argumentative and Linguistic Patterns of LLMs in Persuasive Contexts
- Spontaneous Persuasion: An Audit of Model Persuasiveness in Everyday Conversations
Original note title
LLM-generated arguments require higher cognitive effort than human-generated arguments yet match their persuasive force